bszeti/openshift-grafana-dashboards

openshift-grafana-dashboards's Issues

Inclusion of terminated pods in calculations

Hi,

I tried out the dashboards and they seem very nice however I found a mistake when examining the dashboard data in our environment. For instance in the "Summary by node" Dashboard, panel "CPU requests/allocatable %" includes pods which have completed and thus could express a percentage > 100%. The query in this case:

sum(    
    kube_pod_container_resource_requests_cpu_cores{node=~"$node"}       
    )
/
sum(    
    kube_node_status_allocatable_cpu_cores{node=~"$node"}       
    )

This could lead you to believe your cluster is more full than it actual is. The only way I have found to fix calculations as such would be to examine pod status and likely means running some Prometheus recording rules. This is because the addition of other labels into that query would be "joining" metrics which is heavy on Prometheus.